[Emergency ultrasound for blunt abdominal trauma--meta-analysis update 2003]

Emergency ultrasound has established itself as a key procedure of primary diagnostic work-up for blunt abdominal and multiple trauma. However, in a systematic review published in 2001 ultrasonography turned out to provide an unexpectedly low sensitivity. We conducted an update of this analysis to investigate if test characteristics will be maintained including recent studies. Prospective trials published between January 1957 and January 2003 were identified using the Medline/Oldmedline, Embase and Cochrane Controlled Trials Register databases. The searching strategy comprised a manual search as well as a search along the world-wide web. Qualitative rating was carried out by two investigators using criteria proposed by the Centre for Evidence-Based Medicine, Oxford. We investigated a composite endpoint (i. e., free fluid and/or organ laceration) as well as the single criteria organ injury and free intraabdominal fluid collections. After calculation of two-by-two-tables, Summary Receiver Operating Characteristics (SROC) and Q* values were determined together with their 95% confidence intervals. The Q* value was proposed as the point of intersection where sensitivity equals specificity. In addition, a random effects model was employed to compute common positive and negative likelihood ratios (LR). By assessing the title and/or abstract, 349 of 957 papers contained potentially valid information for the purpose of this review. A total of 67 studies were deemed eligible, nine of which had to be excluded from meta-analysis because of dual publication. This left 58 trials allocating 16,361 subjects for statistical analysis. Despite a trend towards improved study designs observed during the past decade, the included trials were of average methodological quality. Two-thirds of all investigations fulfilled two or less of the six possible quality criteria. The diagnostic reference standard was applied independently in only 40% of all protocols. With regard to the composite endpoint and the sonographic depiction of free fluid, the Q* value was estimated at 0.91, whereas Q* equaled 0.90 for the detection of organ injury. Q* values subsequently decreased with improving study quality and fell clearly below 0.80 in methodologically proper studies. Accounting for a negative LR of 0.23 (composite endpoint) and an assumed prevalence of 35% of intraabdominal injury, a post-test probability of 11% will remain in case of a negative sonogram. In pediatric trauma, ultrasound showed even worse test characteristics (negative LR = 0.43). Thus, in case of a 35% prevalence, the post-test probability has to estimated at 19%. Emergency ultrasound provides high specificity but insufficient sensitivity to reliably rule out intraabdominal injury.